I. Camera Trap Data

SPECIES RICHNESS

Entire Study Area

Excluding arboreal mammals, dogs, and taxa not identified to species level (Muntiacus spp., Tragulus spp. and Unid civets). All camera trap locations: 36 total species.

##   SpeciesRichness
## 1              36

Compared to only CT locations that have LiDAR data: 34 total species

##   SpeciesRichness
## 1              34

By forest type

Excluding arboreal mammals

## # A tibble: 7 × 4
##   habitat           n.scanned n.all  diff
##   <ord>                 <int> <int> <int>
## 1 Montane                  22    29     7
## 2 Upland Granite           22    24     2
## 3 Lowland Granite          20    22     2
## 4 Lowland Sandstone        24    27     3
## 5 Alluvial Bench           27    31     4
## 6 Freshwater Swamp         28    31     3
## 7 Peat Swamp               20    23     3

By partition

Excluding unid animals

## # A tibble: 11 × 4
##    partition n.scanned n.all  diff
##    <ord>         <int> <int> <int>
##  1 MO2              22    26     4
##  2 UG2              22    23     1
##  3 UG1              14    24    10
##  4 LG2              20    21     1
##  5 LG1              18    20     2
##  6 LS2              19    25     6
##  7 LS1              23    26     3
##  8 AB2              22    27     5
##  9 AB1              21    24     3
## 10 FS1              28    31     3
## 11 PS1              20    23     3

Again, but for terrestrial mammals only

By camera trap location

Excluding arboreal mammals

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   1.000   6.000  10.000   9.659  13.000  19.000

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    2.00   10.00   13.00   12.21   15.75   19.00

SHANNON DIVERSITY

By camera trap location

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##   0.000   1.339   1.634   1.589   1.988   2.484

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##  0.3046  1.4367  1.8424  1.7670  2.1257  2.4841

By partition

Comparing Shannon diversity aggregated at partition for all ct locations to only locations with LiDAR data

##    partition shannon_all shannon_scan        diff
## 1        PS1    1.668766     1.760385 -0.09161834
## 2        FS1    2.189094     2.100386  0.08870785
## 3        AB1    2.459775     2.435128  0.02464764
## 4        AB2    2.251804     2.282600 -0.03079550
## 5        LS1    2.417306     2.332208  0.08509861
## 6        LS2    2.245537     2.216715  0.02882264
## 7        LG1    2.130200     2.106380  0.02382062
## 8        LG2    2.301889     2.269269  0.03262019
## 9        UG1    1.996149     1.872830  0.12331926
## 10       UG2    2.399767     2.330706  0.06906144
## 11       MO1    2.373866     2.429092 -0.05522585
## 12       MO2    2.340188     2.239893  0.10029504

Creating Shannon diversity metrics for terrestrial mammals only, for all partitions and partitions with montane aggregate

By forest type

##             habitat shannon_all shannon_scan        diff
## 1        Peat Swamp    1.668766     1.760385 -0.09161834
## 2  Freshwater Swamp    2.189094     2.100386  0.08870785
## 3    Alluvial Bench    2.380981     2.411738 -0.03075691
## 4 Lowland Sandstone    2.366832     2.337795  0.02903769
## 5   Lowland Granite    2.247289     2.235305  0.01198368
## 6    Upland Granite    2.265847     2.173927  0.09191952
## 7           Montane    2.425421     2.385883  0.03953851

SPECIES EVENNESS

By camera trap location

Pielou’s diversity measure of species evenness

By partition

Pielou’s diversity measure of species evenness

adding diversity metrics to forest structure metrics df

FUNCTIONAL DIVERSITY METRICS

By camera trap location

## Species x species distance matrix was not Euclidean. 'sqrt' correction was applied. 
## FEVe: Could not be calculated for communities with <3 functionally singular species. 
## FDis: Equals 0 in communities with only one functionally singular species. 
## FRic: To respect s > t, FRic could not be calculated for communities with <3 functionally singular species. 
## FRic: Dimensionality reduction was required. The last 32 PCoA axes (out of 34 in total) were removed. 
## FRic: Quality of the reduced-space representation (based on corrected distance matrix) = 0.5079213 
## FDiv: Could not be calculated for communities with <3 functionally singular species.

By partition

## Species x species distance matrix was not Euclidean. 'sqrt' correction was applied. 
## FRic: Dimensionality reduction was required. The last 16 PCoA axes (out of 34 in total) were removed. 
## FRic: Quality of the reduced-space representation (based on corrected distance matrix) = 0.9930462

Correlation between community metrics

By camera trap location:

By partition:

COMMUNITY METRICS VISUALIZATION

Community metrics by CT location/forest type

Community metrics by elevation

DISTANCE x DISSIMILARITY

Calculating pairwise distance for scanned locations and plotting values by pairwise similarity, with blue points showing pairs in the same forest type and red points pairs in different forest types. I included the six community metrics here as well, but it might be worthwhile to see what the linear models look like for these for the full set of camera traps, since we know that the community metrics for only the scanned sites differs from the values for all CT sites, especially in the higher elevation forest types.

Histogram of pairwise distances between scanned locations to see if there are any obvious breaks here that can guide grid size

SPATIAL AUTOCORRELATION

Firdt I want to just visualize spatial patterns of structure variables across the study site, by camera trap location.

Global Moran’s I

A formal test of spatial autocorrelation across the entire study area. I may only be able to meaningfully test for spatial autocorrelation at this scale, and not at the grid or partition scales, as the typical low end of spatial units required for reliable statistical inference is 30.

Five structure metrics show a significant level of spatial autocorrelation, including max height, vertical rumple, vegetation volume, leaf area density, and mean tree height.

Looking at the spatial patterns for these metrics in the plots above, max height, vrumple, lad, and mean height seem to show a roughly similar pattern, with low values in the swamp forests and montane, and highest values in the lowland forests in the center of the trail system - exactly what we see in the boxplots of these values plotted by forest type. Vegetation volume (vFRcanopy) shows a different spatial pattern. And rumple, along with closedgapspace show an outlier in the center of the trail system - location BC 12.4 (not sure what’s going on with that one)

## # A tibble: 21 × 7
##    Metric          Moran_I Expected_I Variance_I  Z_score    P_value Significant
##    <chr>             <dbl>      <dbl>      <dbl>    <dbl>      <dbl> <lgl>      
##  1 FDiver          0.416      -0.0179    0.00628  5.48       2.16e-8 TRUE       
##  2 max.height      0.417      -0.0175    0.00642  5.42       3.01e-8 TRUE       
##  3 vzrumple        0.299      -0.0175    0.00634  3.98       3.43e-5 TRUE       
##  4 shannon_ct      0.293      -0.0175    0.00631  3.91       4.66e-5 TRUE       
##  5 vFRcanopy       0.230      -0.0175    0.00623  3.13       8.64e-4 TRUE       
##  6 lad.max         0.156      -0.0175    0.00640  2.17       1.51e-2 TRUE       
##  7 mean.tree.h     0.146      -0.0175    0.00629  2.06       1.98e-2 TRUE       
##  8 n.all           0.121      -0.0175    0.00640  1.73       4.20e-2 TRUE       
##  9 div_even        0.116      -0.0175    0.00640  1.67       4.76e-2 TRUE       
## 10 sd.r            0.0916     -0.0175    0.00643  1.36       8.67e-2 FALSE      
## 11 pts.below.2m    0.0885     -0.0175    0.00629  1.34       9.06e-2 FALSE      
## 12 FRich           0.0825     -0.0179    0.00629  1.27       1.03e-1 FALSE      
## 13 stand.dens      0.0804     -0.0175    0.00645  1.22       1.11e-1 FALSE      
## 14 stem.vol        0.0130     -0.0175    0.00623  0.387      3.49e-1 FALSE      
## 15 rumple         -0.00832    -0.0175    0.00470  0.135      4.46e-1 FALSE      
## 16 CRR.rho        -0.0163     -0.0175    0.00641  0.0151     4.94e-1 FALSE      
## 17 ClosedGapSpace -0.0183     -0.0175    0.00549 -0.00964    5.04e-1 FALSE      
## 18 basal.area     -0.0286     -0.0175    0.00624 -0.140      5.56e-1 FALSE      
## 19 zentropy       -0.0543     -0.0175    0.00625 -0.465      6.79e-1 FALSE      
## 20 FEven          -0.0883     -0.0179    0.00639 -0.880      8.11e-1 FALSE      
## 21 mean.dbh       -0.105      -0.0175    0.00639 -1.09       8.63e-1 FALSE

Extracting Moran’s eigenvector values to use in glm to control for spatial autocorrelation

CREATING SURVEY GRIDS

Run grid size functions

MODELING - Terrestrial mammals (CT data)

Descriptions of forest structure metrics used in models

I used {FORTLS} to extract the following metrics:

  1. max.height - maximum height of trees in the stand, in m

  2. mean.tree.h - mean tree height in the stand, in m.

  3. CRR.rho - canopy relief ratio, using rho for horizontal distance. A measure of canopy variation, with lower scores indicating lower local variation in canopy surface, i.e., more uniform age canopy/ less canopy surface complexity.

  4. sd.r - standard deviation of tree heights in stand in m. I’ve seen this value used as an approximation for vertical stratification.

  5. mean.dbh - mean diameter at breast height of trees in stand, in cm (measured at height of 1.3m)

  6. basal.area - basal area of trees in the stand (m2/ha)

  7. stand.dens - density of trees in the stand (trees/ha)

  8. stem.vol - volume of trees in the stand (m3/ha)

  9. pts.below.2m - shows the number of points in the point cloud that are below 2m.

I used {lidr} and {lidRmetrics} to extract the following metrics:

  1. zentropy - normalized Shannon diversity index of z (height) values. Describes vertical complexity.

  2. lad.max - Leaf Area Density maximum value for 1m vertical bins. Describes maximum foliage cover of the point cloud.

  3. rumple - Rumple index (rugosity), a ratio of the canopy surface area to it’s projected ground area.

vn - number of 1m voxels created for the following volumetric metrics.

  1. vFRcanopy - ratio of filled to empty voxels, only counting cells within and below canopy, ignoring above.

  2. vzrumple - vertical Rumple index.

  3. ClosedGapSpace - volume of voxels that are classified as gaps underneath the canopy

Check correlation between predictors.

Basal area is highly correlated (>0.8) with stem volume and stand density.

Models by camera trap location

Fitting global model then using {dredge} to find models of best fit. Restricting models to five predictors to avoid overfitting. Excluding two highly correlated pairs in dredge models: basal area & stand density, and basal area & stem volume.

Species richness model

averaged model output - averaging all models

Shannon diversity model

averaged model output - averaging all models

Species evenness model

averaged model output - averaging all models

Functional richness model

Top model summary output

averaged model output - averaging all models

Functional evenness model

Top model summary output

averaged model output - averaging all models

Functional divergence model

Top model summary output

averaged model output - averaging all models

Comparing coefficient plot from averaged models

Plotting variables from models that are at least closely reliable predictors (50% CI’s don’t overlap zero).

MODELS BY GRID

75 ha grid

The 75 ha grid creates 17 groups, with most groups (grid cells) containing 2 - 4 scanned locations each.

Correlation of community metrics. Richness is highly correlated with FRich and FDiver, and shannon is highly correlated with evenness. So probably should only keep shannon and the functional metrics.

Correlation of structure metrics - basal area and stand dens, basal area and stem vol, %<2m and sd.r, max h and mean h, max h and vrumple, and mean h and vrumple

Species richness model

Fitting global model then using {dredge} to find models of best fit.

Top model summary output. Restricting models to two predictors to avoid overfitting

averaged model output - averaging all models

Species diversity model

averaged model output - averaging all models

Species eveness model

averaged model output - averaging all models

Functional richness model

averaged model output - averaging all models

Functional evenness model

averaged model output - averaging all models

Functional divergence model

averaged model output - averaging all models

Comparing coefficient plots from averaged models

Only for shannon diveristy and functional models, based on high degree of correlation with the other community metrics

170 ha grid

The 170 ha grid creates 10 groups (grid cells)

Correlation of community metrics. Richness is highly correlated with FRich (0.87) and shannon is highly correlated with evenness (0.93). So probably should only keep shannon and the functional metrics.

Correlation of structure metrics - basal area and stand dens, basal area and mean dbh, max h and mean h, max h and vrumple, rumple and lad,

Species richness model

averaged model output - averaging all models

Species diversity model

averaged model output - averaging all models

Species eveness model

averaged model output - averaging all models

Functional richness model

The beta coefficients are very large here, not sure why this is so much different from the rest of the models. Maybe an odd convergence issue?

averaged model output - averaging all models

Functional evenness model

averaged model output - averaging all models

Functional divergence model

averaged model output - averaging all models

Comparing coefficient plots from averaged models

Showing only richness, shannon, FEven, and FDiver models, since FRich seemingly had nonsense results

MODELS BY PARTITION

Check correlation between predictors.

Basal area is highly correlated (>0.8) with stand density, max height with mean tree h and vzrumple, and vzrumple with mean tree height and rumple.

Species richness model

Fitting global model then using {dredge} to find models of best fit.

Top model summary output. Restricting models to two predictors to avoid overfitting

averaged model output - averaging all models

Species diversity model

Top model summary output.

averaged model output - averaging all models

Species evenness model

Top model summary output.

averaged model output - averaging all models

Functional richness model

The model output here seems odd. After looking at the fric values closer, there seems to be an outlier here which may be throwing off the analysis. Not sure what to do here other than not include it in the final analysis, which may be warranted based on the high correlation between the richness and fric values (0.81)

averaged model output - averaging all models

Functional evenness model

Top model summary output.

averaged model output - averaging all models

Functional divergence model

Top model summary output.

averaged model output - averaging all models

Comparing coefficient plot from averaged models, reliable predictors only

Since fric is highly correlated with richness (0.81) it makes sense to ignore this model here. Of the other models that have reliable predictors, the shannon and evenness values are also highly correlated (0.89) and both show vFRcanopy (vegetated volume) as the sole reliable predictor, so I also omit the evenness model here.

Comparing avg model multiplots for CT and partition scales

ORDINATION ANALYSIS

Using PCA to see how scan locations group together and to identify which forest structure metrics best differentiate between forest types and partitions.

Which structure metrics best differentiate between forest types?

Table of variable PCA contribution values

## # A tibble: 15 × 2
##    variable       total_contrib
##    <chr>                  <dbl>
##  1 basal.area             89.6 
##  2 mean.tree.h            88.4 
##  3 max.height             79.5 
##  4 vzrumple               68.6 
##  5 stem.vol               68.3 
##  6 rumple                 65.1 
##  7 stand.dens             60.8 
##  8 ClosedGapSpace         56.5 
##  9 pts.below.2m           46.9 
## 10 mean.dbh               41.9 
## 11 sd.r                   41.3 
## 12 lad.max                31.9 
## 13 zentropy               31.1 
## 14 vFRcanopy              27.2 
## 15 CRR.rho                 2.79

BIVARIATE PLOTS

terrestrial mammals (ct data) by ct location

Structure metrics X species richness

Structure metrics X Shannon diversity

Structure metrics X Species Evenness

Structure metrics X Functional Richness

Structure metrics X Functional Evenness

Structure metrics X Functional Divergence

terrestrial mammals (ct data) by partition

Structure metrics X species richness

Structure metrics X Shannon diversity

Structure metrics X Species Evenness

summary table for Pearson correlation between structure metrics and diversity metrics
Pearson correlations beteen structure metrics and diversity metrics
Camera Trap Scale
Partition Scale
Richness Shannon Diversity Evenness Func. Richness Func. Evenness Func. Divergence Richness Shannon Diversity Evenness Func. Richness Func. Evenness Func. Divergence
max.height -0.2822665 -0.3473963 -0.1104631 -0.3501106 -0.0092927 -0.0180895 -0.0606541 -0.0892353 -0.0616092 -0.0299001 -0.6616431 0.2567726
sd.r 0.1366387 0.1075038 -0.0524271 0.0512050 -0.2323421 -0.0934397 0.4907328 0.2844402 0.1164183 0.3034671 -0.3208231 0.4682592
CRR.rho -0.1384227 -0.1656814 -0.1351897 0.0590025 0.0455121 -0.0259802 -0.2350246 -0.2878944 -0.1892790 -0.0155821 -0.2928088 0.3342937
pts.below.2m -0.0373671 -0.1858660 -0.2586856 -0.1331214 -0.1344263 -0.1652155 0.7039527 0.1522458 -0.0934773 0.5010931 -0.2090541 -0.0752225
stand.dens 0.3639998 0.3503880 0.0094444 0.3978090 -0.2071896 0.0289664 0.2540969 0.3944726 0.3157581 0.2429591 0.3659963 0.1044852
basal.area 0.2649925 0.3049558 0.0734725 0.2210968 -0.0917111 -0.0339108 0.0731401 0.4398316 0.4334561 0.2395054 0.1886525 0.0451641
stem.vol 0.2552954 0.3036081 0.1124067 0.1919080 -0.0063824 0.1078632 -0.1502708 0.4492890 0.5219881 0.0616719 0.0387210 0.5341543
mean.tree.h -0.0877811 -0.1066189 -0.0080111 -0.2599512 -0.0765966 0.0291181 -0.1133321 -0.0552613 -0.0043525 -0.0587486 -0.6701856 0.2090019
mean.dbh -0.0111331 0.0211614 0.0084297 -0.1317456 0.0534071 -0.0689409 -0.2418334 0.1612925 0.2746417 0.1527113 -0.3235092 -0.1038412
zentropy 0.1625037 0.2713834 0.1940172 0.1645552 -0.0766785 0.2367055 -0.1977574 0.1336686 0.2014013 -0.2096010 0.5166602 0.0718520
lad.max 0.0900443 0.1415487 0.1229312 0.1186556 -0.2089666 0.0281173 0.0236891 0.3488180 0.3407154 0.0597146 -0.3630760 0.4903437
rumple 0.1010440 0.0289683 -0.0072435 -0.0817517 0.0184174 0.0902568 -0.0510142 0.2803597 0.2732869 -0.2904670 -0.2530193 0.0537974
vFRcanopy 0.2868723 0.2197312 -0.0868493 0.2935773 0.1033008 0.0374483 0.0678703 0.3658665 0.3394809 0.1350511 0.6458722 -0.0900869
vzrumple -0.0764892 -0.2248555 -0.2254738 -0.1783032 0.0698351 -0.0816645 -0.1178366 -0.0060737 0.0383544 -0.0418600 -0.6232796 0.2198470
ClosedGapSpace 0.0262216 0.0010171 0.0707572 -0.1061820 -0.0547225 0.1677327 -0.0621519 0.1262294 0.1124200 -0.4682253 -0.1449529 -0.1121964